Article Text

Download PDFPDF
Original article
Non-invasive imaging to identify susceptibility for ventricular arrhythmias in ischaemic left ventricular dysfunction
  1. Mischa T Rijnierse1,
  2. Cornelis P Allaart1,
  3. Stefan de Haan1,
  4. Hendrik J Harms2,
  5. Marc C Huisman2,
  6. Aernout M Beek1,
  7. Adriaan A Lammertsma2,
  8. Albert C van Rossum1,
  9. Paul Knaapen1
  1. 1Department of Cardiology, Institute for Cardiovascular Research (ICaR-VU), VU University Medical Center, Amsterdam, the Netherlands
  2. 2Department of Radiology and Nuclear Medicine, VU University Medical Center, Amsterdam, the Netherlands
  1. Correspondence to Dr P Knaapen, VU University Medical Center, Department of Cardiology, De Boelelaan 1117, P.O. Box 7057, Amsterdam 1081 HV, the Netherlands; p.knaapen{at}


Objective Non-invasive imaging of myocardial perfusion, sympathetic denervation and scar size contribute to enhanced risk prediction of ventricular arrhythmias (VA). Some of these imaging parameters, however, may be intertwined as they are based on similar pathophysiology. The aim of this study was to assess the predictive role of myocardial perfusion, sympathetic denervation and scar size on the inducibility of VA in patients with ischaemic cardiomyopathy in a head-to-head fashion.

Methods 52 patients with ischaemic heart disease and left ventricular ejection fraction (LVEF) ≤35%, referred for primary prevention implantable cardioverter-defibrillator (ICD) implantation, were included. Late gadolinium-enhanced cardiovascular MRI was performed to assess LV volumes, function and scar size. Using [15O]H2O and [11C]hydroxyephedrine positron emission tomography, both resting and hyperaemic myocardial blood flow (MBF), and sympathetic innervation were assessed. After ICD implantation, an electrophysiological study (EPS) was performed and was considered positive in case of sustained VA.

Results Patients with a positive EPS (n=25) showed more severely impaired global hyperaemic MBF (p=0.003), larger sympathetic denervation size (p=0.048) and tended to have larger scar size (p=0.07) and perfusion defect size (p=0.06) compared with EPS-negative patients (n=27). No differences were observed in LV volumes, LVEF and innervation-perfusion mismatch size. Multivariable analysis revealed that impaired hyperaemic MBF was the single best independent predictor for VA inducibility (OR 0.78, 95% CI 0.65 to 0.94, p=0.007). A combination of risk markers did not yield incremental predictive value over hyperaemic MBF alone.

Conclusions Of all previously validated approaches to evaluate the arrhythmic substrate, global impaired hyperaemic MBF was the only independent predictor of VA inducibility. Moreover, a combined approach of different imaging variables did not have incremental value.

Statistics from

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.


Prediction of ventricular arrhythmias (VA) is essential to identify patients at highest risk for sudden cardiac death (SCD). Currently, impaired left ventricular ejection fraction (LVEF) is the main selection criterion for eligibility of implantable cardioverter-defibrillator (ICD) therapy for primary prevention of SCD.1 Nonetheless, several long-term follow-up studies report an ICD discharge rate of only 11–35% among patients with ICDs for primary prevention,2 ,3 whereas the majority of SCD cases occur in patients with preserved LVEF.4 Multiple previous studies have investigated novel risk markers that assess the substrate of VA using different cardiac imaging modalities.5 In patients with ischaemic left ventricular dysfunction, the characterisation of scar tissue using late gadolinium-enhanced cardiovascular MRI (LGE-CMR) has been frequently linked to VA inducibility and SCD.6–9 More recently, nuclear imaging studies have demonstrated that evaluating the extent of denervation, which often exceeds the infarct size, might enhance risk prediction for VA.10–12 In addition, denervated yet viable myocardium may in particular be prone to VA.12 ,13 Moreover, residual ischaemia even after revascularisation may serve as an important arrhythmic substrate.14 ,15 Some of these different approaches to evaluate the arrhythmic substrates are, however, intertwined and may actually represent similar pathophysiology. Therefore, comparison of several risk markers within a single patient population is warranted to define the best strategy to select patients that are at highest risk for SCD. The current study is a head-to-head comparison to investigate the role of positron emission tomography (PET)-assessed myocardial perfusion and sympathetic denervation, as well as LGE-CMR-assessed scar size in predicting VA inducibility in patients with ischaemic heart disease and left ventricular dysfunction.


Study population

Patients with ischaemic heart disease and LVEF ≤35% who were referred for ICD implantation for primary prevention of SCD according to current guidelines1 were prospectively included. Exclusion criteria were history of documented sustained VA, no sinus rhythm and contraindications for CMR, PET or an electrophysiological study (EPS).

PET protocol

All PET studies were performed using a PET/CT device (Philips Gemini TF 64, Philips Healthcare, Best, the Netherlands). Patients were instructed to refrain from intake of products containing caffeine or xanthine 24 h prior to the scan and all patients received a radial artery catheter for manual arterial blood sampling. The PET imaging protocol was described in detail previously.16 First, a 6 min dynamic [15O]H2O PET scan was performed to assess myocardial blood flow (MBF) under resting conditions, immediately followed by a respiration-averaged low-dose CT scan to correct for attenuation. After an interval of 10 min, a second identical [15O]H2O PET sequence was performed under infusion of intravenous adenosine (140/µg/kg//min), which was started 2 min prior to the start of the dynamic stress scan and terminated after the low-dose CT. Dynamic [15O]H2O images were reconstructed into 22 frames (1×10, 8×5, 4×10, 2×15, 3×20, 2×30 and 2×60 s) using the three-dimensional row-action maximum likelihood algorithm (3D RAMLA), applying all appropriate corrections. Finally, after an interval of approximately 15 min, a 60 min dynamic [11C]hydroxyephedrine (HED) PET scan was performed, immediately followed by a low-dose CT scan. During this scan, 7 mL arterial samples were collected manually at 2.5, 5, 10, 20, 30, 40 and 60 min to determine plasma and whole blood activity concentrations, and radiolabelled plasma [11C]HED metabolites. Dynamic [11C]HED images were reconstructed into 36 frames (1×10, 8×5, 4×10, 3×20, 5×30, 5×60, 4×150, 4×300, and 2×600 s) using the 3D RAMLA with application of all appropriate corrections.

PET data analysis

PET data analysis was performed using in-house developed software. For both [15O]H2O and [11C]HED, image-derived input functions were derived by placing 1 cm diameter regions of interest (ROIs) over the ascending aorta in at least five transaxial planes showing the first pass of the injected bolus. These ROIs were combined in one volume of interest (VOI) for the ascending aorta. Furthermore, a right ventricular VOI was obtained by drawing a second set of ROIs in at least three transaxial planes over the right ventricle cavity. Subsequently, both VOIs were transferred to the full dynamic images to obtain arterial whole blood and right ventricular time–activity curves (TAC). For [11C]HED image analysis, parent fractions and ratios of plasma/whole blood concentrations derived from the manual blood samples were fitted to a sigmoid function.17 Subsequently, the arterial whole blood TAC was multiplied by the fitted plasma/whole blood ratio and parent fraction curves. For MBF analysis, parametric images of MBF, anatomic tissue fraction, perfusable tissue fraction (PTF), perfusable tissue index (PTI) and arterial and venous blood volume fractions were generated as previously described.18 ,19 Segmental VOIs were defined manually on short-axis PTF images. Finally, these VOIs were projected onto the entire dynamic [15O]H2O emission scans to extract segmental TACs that were fitted to a single-tissue compartment model. For [11C]HED image analysis, segmental VOIs were drawn on short-axis images in the final frame of the dynamic scan. All segmental VOIs were drawn according to the 17-segment model of the American Heart Association.

MBF was expressed in mL/min/g of perfusable tissue. Coronary flow reserve (CFR) was defined as the ratio of hyperaemic and resting MBF. In addition, resting MBF was multiplied by the PTI to obtain transmural MBF values (MBFT) for perfusion defect size calculation. Myocardial [11C]HED uptake was expressed using the retention index (RI), which was calculated as the uptake at the last frame (50–60 min) divided by the integral of the arterial plasma-corrected TAC. In all patients, a reference area was designated during [15O]H2O and [11C]HED imaging analysis by selecting four myocardial remote segments that showed the highest values of MBFT and [11C]HED RI, respectively. Subsequently, resting perfusion and innervation defects were defined as myocardium with MBFT and [11C]HED RI<50% of the mean MBFT and [11C]HED RI of the reference segments. To evaluate the potential influence of global downregulation of [11C]HED uptake, mean [11C]HED RI in reference segments was calculated. Finally, the mismatch between innervation defect and perfusion defect was calculated as the difference between the [11C]HED defect and MBFT defect.

CMR protocol and image analysis

CMR studies and image analyses were performed as described previously.15 In short, a clinical 1.5-T MRI scanner (Magnetom Avanto; Siemens, Erlangen, Germany) was used with a dedicated phased-array body coil. After survey scans, cine imaging was performed using a retrospectively ECG-gated, steady-state free precession sequence during breath holds in mild expiration. Standard four-chamber, three-chamber and two-chamber orientations were obtained, and subsequently, a stack of 10–12 consecutive short-axis slices was acquired, fully covering the left ventricle. Approximately 10–15 min after administration of 0.2 mmol/kg gadolinium, LGE images were acquired in similar orientations as the cine images using a two-dimensional segmented inversion-recovery prepared gradient echo sequence. In case of difficulties with breath holding during LGE imaging, a single-shot sequence was used instead of a segmented sequence.

Images were analysed using the dedicated software package MASS (Mass V.5.1 2010-EXP beta, Medis, Leiden, the Netherlands). Endocardial and epicardial borders of short-axis cine images were outlined manually in both end-diastolic and end-systolic phases to quantify left ventricular volumes, LVEF and mass. Subsequently, endocardial and epicardial contours of the short-axis LGE images were traced manually. Myocardial scar size was automatically quantified using the full-width at half-maximum method, which defines scar as myocardium with signal intensity of ≥50% of the maximum signal intensity in the hyperenhanced area.7 Total scar size was expressed as grams and percentages of the total left ventricular mass.

Electrophysiological study

A non-invasive EPS was performed after ICD implantation using the device programmer and consisted of a programmed stimulation protocol including two 8-beat drive trains (cycle lengths of 600 and 400 ms) immediately followed by up to three extrastimuli.15 Extrastimuli were delivered at progressively shorter coupling intervals in 20 ms steps until the effective refractory period or 180 ms was reached. If no sustained VA was induced, the protocol was repeated with a basic cycle length of 400 ms under infusion of isoproterenol (aimed baseline heart rate 100–110/min). The EPS was performed and interpreted by an experienced electrophysiologist, blinded to the imaging data. A positive EPS was defined as the induction of a sustained monomorphic or polymorphic VA lasting >30 s or requiring termination in case of haemodynamic instability.

Statistical analysis

Continuous variables are presented as mean±SD, and categorical data are summarised as frequencies and percentages. Histograms were used to evaluate whether continuous data were normally distributed. Groups with positive and negative EPS were compared using the independent Student's t test for continuous variables and two-tailed Fisher's exact test for categorical data. Levene’s test for equality of variances was used to verify whether the independent Student’s t test was appropriate. Non-normal data were compared with Mann–Whitney U test when appropriate. Comparisons of paired data were performed with the paired t test. Receiver operating characteristic (ROC) curve analysis was performed and areas under the curve (AUC) were calculated. Furthermore, univariable logistic regression analyses were performed for all clinical and imaging variables that may be associated with VAs. Variables with p value <0.10 at univariable analysis were entered in a multivariable model using a forward stepwise selection procedure to prevent overfitting. The following variables were entered in the multivariable model: hyperaemic MBF, resting perfusion defect size, innervation defect size and ACE/angiotensin II receptor blockers (ARBs) usage. Furthermore, multivariable logistic regression analyses were repeated, each time entering the most significant predictor of a positive EPS (hyperaemic MBF), combined with other imaging variables to investigate their interrelations and combined predictive power. The predictive power of each multivariable model was expressed by the Nagelkerke R2 and ROC AUC analysis. All tests were performed two-sided and were considered statistically significant if p value was <0.05. All statistical analyses were performed using SPSS software package (SPSS V.20.0, IBM Corporation, Chicago, Illinois, USA).


Study population

Fifty-two patients with ischaemic left ventricular dysfunction who underwent an EPS after ICD implantation were included. Clinical baseline characteristics are presented in table 1, and figure 1 displays some graphical examples of the imaging protocol. Sustained VA was inducible in 25 (48%) patients. Most of the patients (n=22) with a positive EPS showed a monomorphic VA, whereas three patients showed a polymorphic VA. Isoproterenol infusion was not needed to induce VA in the majority of patients with a positive EPS (n=22), while in three patients a monomorphic VA was induced during infusion of isoproterenol. Patients with a positive EPS used significantly less often ARB or ACE inhibitors compared with patients having a negative EPS (68% vs 93%, respectively, p=0.04). Among patients who underwent percutaneous coronary intervention/coronary arterial bypass graft surgery, the revascularisation procedure was successful in 98%, which did not differ significantly between EPS-positive and EPS-negative patients (95% vs 100%, respectively, p=0.49). None of the patients were considered candidates for (further) revascularisation. No other differences in baseline characteristics were observed between both groups.

Table 1

Baseline characteristics

Figure 1

Examples of positron emission tomography (PET) and late gadolinium enhanced-cardiovascular MRI (LGE)-CMR results in two patients with electrophysiological study (EPS)-positive (A) and EPS-negative (B) result. (A) A patient with LGE scar in the inferolateral wall with corresponding PET perfusion and innervation defects, and significant innervation-perfusion mismatch as depicted in red. Quantitative PET revealed a severely impaired hyperaemic myocardial blood flow (MBF). (B) A patient with LGE scar in the anteroseptal wall with corresponding PET perfusion and innervation defects, and limited area of innervation-perfusion mismatch (red area). Quantitative PET revealed a relatively preserved hyperaemic MBF.

Myocardial perfusion and sympathetic denervation

Table 2 lists the PET results. Due to technical issues, PET imaging was not performed in three patients and another two patients underwent an incomplete PET protocol (in one patient resting MBF was not obtained due to severe motion artefacts, in another patient HED imaging was missing due to tracer production issues). As demonstrated in figure 2A, no differences were observed in global resting MBF and CFR between patients with positive and negative EPS results. However, patients with inducible VA during EPS showed significantly lower hyperaemic MBF (1.36±0.39 mL/min/g vs 1.71±0.39 mL/min/g, p=0.003). The global innervation defect size significantly exceeded the global perfusion defect size (24.2%±12.6% vs 18.3%±12.0%, respectively, p<0.001). Figure 2B presents global perfusion defect, innervation defect and innervation-perfusion mismatch compared between EPS-positive and EPS-negative groups. A larger area of denervation was observed in patients who showed VA inducibility as compared with negative EPS patients (27.9%±11.9% vs 20.7%±12.6%, p=0.048). Furthermore, a trend was observed towards a larger perfusion defect in patients with inducible VA during EPS (21.8%±12.8% vs 15.3%±10.7%, p=0.06). The mismatch between the innervation defect and perfusion defect, however, did not differ between both groups (6.2%±7.5% vs 5.9%±8.0%, p=0.88). No difference was observed in mean [11C]HED RI in reference segments between EPS-positive and EPS-negative patients (3.40±0.93 vs 3.73±0.94, respectively, p=0.23).

Figure 2

Aligned scatter plot comparisons of PET-assessed quantitative global perfusion (A), innervation and perfusion defect sizes (B) and LGE-CMR assessed scar size (C) between patients with negative and positive electrophysiological study (EPS). CFR, coronary flow reserve; MBF, myocardial blood flow.

Table 2

PET results

Myocardial function and infarct size

Table 3 presents LGE-CMR results. In one patient, LGE could not be interpreted due to poor image quality. No differences were observed in LVEF, LV volumes and mass between patients with positive and negative EPS. Patients with VA inducibility tended to have larger scar size as quantified in grams (23.8±13.1 g vs 18.1±9.2 g, p=0.07). The difference in scar size expressed as percentage of the total LV mass between EPS-positive and EPS-negative patient groups did not reach significance either (18.6%±8.7% vs 14.7%±8.2%, respectively, p=0.10) (figure 2C).

Table 3

CMR results

Predictors for ventricular arrhythmia inducibility

Univariable and multivariable analysis of clinical and imaging parameters for predicting VA inducibility is presented in table 4. Parameters that were significantly associated with VA inducibility in univariable analysis included usage of ACE/ARB (OR 0.17, 95% CI 0.03 to 0.90, p=0.04) and hyperaemic MBF (OR 0.79, 95% CI 0.66 to 0.94, p=0.007). Multivariable analysis using forward stepwise addition of variables revealed that hyperaemic MBF was the single independent predictor for VA inducibility (OR 0.78, 95% CI 0.65 to 0.94, p=0.007). Figure 3 demonstrates the ROC curves for predicting VA inducibility. Hyperaemic MBF yielded the highest predictive value (AUC 0.74, 95% CI 0.60 to 0.88, p=0.004), whereas the AUC for scar size (AUC 0.62, 95% CI 0.49 to 0.79, p=0.14), innervation defect size (AUC 0.64, 95% CI 0.49 to 0.80, p=0.09) and perfusion defect size (AUC 0.66, 95% CI 0.51 to 0.82, p=0.052) were weak predictors and did not significantly diverge from the line of identity. Optimal cut-off value for hyperaemic MBF derived from the Youden index was <1.37 mL/min/g, which yielded a sensitivity of 61% and specificity of 81% for predicting VA inducibility (positive predictive value 74%, negative predictive value 70%). Table 5 presents four multiple multivariable models, each including a combination of two imaging variables. In every model, hyperaemic MBF predicted VA inducibility independently of scar, perfusion defect and innervation defect with similar ORs, ranging from 0.78 to 0.80. All multivariable models showed comparable predictive value (R2: 0.27–0.31). The AUC of hyperaemic MBF combined with scar size (0.76, 95% CI 0.62 to 0.89, p=0.003), perfusion defect (0.78, 95% CI 0.65 to 0.91, p=0.001) or innervation defect (0.78, 95% CI 0.65 to 0.91, p=0.001) did not yield significant incremental value over hyperaemic MBF as a single predictor.

Table 4

Univariable and multivariable forward stepwise logistic analysis of clinical and imaging parameters for predicting ventricular arrhythmia inducibility

Table 5

Multivariable logistic and ROC analyses of multiple imaging parameters for predicting ventricular arrhythmia inducibility

Figure 3

Receiver operating characteristic (ROC) curve analyses of positron emission tomography-assessed hyperaemic myocardial blood flow (MBF), innervation defect and perfusion defect for predicting inducibility of ventricular arrhythmia. AUC, area under the curve.


The current study evaluated the role of PET-assessed myocardial perfusion, sympathetic denervation, innervation-perfusion mismatch and LGE-CMR-assessed scar size in predicting VA inducibility in patients with ischaemic left ventricular dysfunction. Patients with inducible VA showed more severely impaired hyperaemic perfusion, larger areas of cardiac sympathetic denervation and tended to have larger scar size, whereas no differences were observed in innervation-perfusion mismatch size. Of all used approaches to evaluate the potential arrhythmic substrate by PET and CMR imaging, impairment in quantitatively assessed global hyperaemic perfusion showed highest predictive value. Moreover, it was independent of other imaging parameters and patient characteristics. Furthermore, a combined approach of different imaging modalities did not yield incremental value over the sole assessment of hyperaemic perfusion in selecting patients that may be at highest risk of VA.

Myocardial perfusion

Ischaemia is an important trigger or substrate for VA.14 ,20 ,21 Consequently, optimal revascularisation strategies are crucial in patients with ischaemic left ventricular dysfunction and are generally performed prior to ICD implantation.1 The current study demonstrated that in a population referred for primary prevention ICD therapy, impairment in global hyperaemic perfusion was related to VA inducibility. These data confirm the results that were obtained in a comparable previous pilot study.15 Quantitative [15O]H2O PET assessment of global hyperaemic MBF in ischaemic left ventricular dysfunction reflect a mixture of residual subtle ischaemia, flow within areas surrounding scar, and flow in remote myocardium. Nonetheless, we have previously demonstrated that regional assessment of hyperaemic MBF within scar and remote areas, or heterogeneity, were all related with VA inducibility but did not have incremental value over the assessment of global hyperaemic MBF.15 The current study, therefore, only evaluated global hyperaemic MBF and showed that the relation of hyperaemic MBF and VA inducibility is independent of scar size or innervation defect size. [15O]H2O PET imaging quantifies MBF in viable tissue that is capable of exchanging water rapidly only; therefore, scar size is excluded in the MBF calculation.22 ,23 In contrast to the previous findings, however, the difference in CFR between EPS-positive and EPS-negative patients did not reach significance. As the CFR is also dependent on resting perfusion, patients with low resting perfusion might have a high CFR, whereas hyperaemic MBF is relatively impaired. Several explanations may account for the observed relation between global impairment in hyperaemic MBF and susceptibility for VA rather than regional damaged myocardium. Although none of the included patients were considered candidates for further revascularisation, residual perfusion abnormalities not amenable for revascularisation may still be present that can be detected and quantified by PET perfusion imaging. These perfusion abnormalities could be caused by more severe diffuse epicardial coronary artery disease or (micro)vascular dysfunction extending beyond epicardial coronary arteries. These results suggest that the severity of perfusion abnormalities, even after revascularisation, may result in (subtle) residual ischaemia that relates to susceptibility for VA.

Sympathetic denervation

The area of sympathetic denervation was found to exceed the infarct size, which has been observed previously and most likely is caused by the higher vulnerability of cardiac nerves to oxygen deprivation than myocytes.24 ,25 The recently published PAREPET study was the first to evaluate the role of PET-assessed myocardial sympathetic innervation, viability and resting perfusion in predicting the occurrence of appropriate ICD therapy in patients with ischaemic left ventricular dysfunction.12 Although the current study used a different definition of denervated myocardium than the PAREPET study (<50% of remote vs <75% of peak, respectively), the observed total innervation defect size was comparable (24%±13% vs 27%±11%, respectively), while both studies showed a total infarct size of 17%±9% and 20%±9%, respectively, resulting in a comparable size of denervated but viable myocardium. The current study showed that the total size of sympathetic denervation was larger in patients that displayed sustained VA during EPS. As areas of sympathetic denervation show altered refractoriness and supersensitivity to sympathetic stimulation causing electrical instability, they might serve as arrhythmic substrate.26 ,27 Consistently, a previous study performed in patients with ischaemic left ventricular dysfunction revealed that patients with positive EPS showed larger areas of late [123I]metaiodobenzylguanidine (MIBG) single-photon emission computed tomography (SPECT) defect size, whereas for perfusion defect size only a trend was observed.10 Moreover, the predictive value of late [123I]MIBG defect was confirmed for appropriate ICD therapy in a more recent study.11 Although the [11C]HED PET defect size is closely related to the early and late [123I]MIBG SPECT defect size, it provides superior spatial resolution and potential for absolute quantification.28 The PAREPET study demonstrated that the total size of cardiac denervation predicted appropriate ICD therapy, independent of LVEF, infarct volume and hibernating myocardium. Consistent with the current study, no differences were found in peak [11C]HED RI in remote myocardium, suggesting that potential downregulation of [11C]HED uptake did not influence these results. However, in the current study the size of denervated myocardium did not predict VA inducibility independently. Furthermore, no relation was observed between VA inducibility and the size of innervation-perfusion mismatch. Although in an animal model promising results were obtained linking the magnitude of the innervation-perfusion mismatch to increased VA inducibility and its location of origin,13 clinical studies have reported inconsequential results.10–12 Studies using SPECT did not find a significant relation between the innervation-perfusion mismatch size and VA inducibility or spontaneous VA,10 ,11 whereas the PAREPET study revealed a significant relation with ICD therapy for fast VA.12 A potential explanation of the inconsistent findings in the current study may be the use of VA inducibility as end point instead of spontaneous fast VA. Second, the spatial resolution of PET imaging may still not be sufficient for a detailed mismatch calculation in every patient. The PAREPET study also showed that global innervation defect was more predictive of VA compared with the mismatch area.12 Finally, technical issues such as patient motion during the extensive scanning protocol with different serial PET scans may have occurred and introduced some difficulties in the mismatch calculation. The results in this study support these findings, suggesting that the total size of denervation might be more importantly related to VA than the mismatch area itself.

Myocardial scar

Interestingly, the current study showed only a trend towards larger scar size in patients with VA inducibility. Bello et al,6 however, revealed that scar size was related to VA inducibility, particularly for monomorphic VA. The current study also considered polymorphic VA as positive EPS result, which might explain the inconsistencies. Furthermore, it has been suggested that the CMR assessment of infarct border zone is more importantly related to VA inducibility.29 However, in a comparable previous pilot study we could not confirm these results, although the included patient population was smaller.15 The relation between scar characteristics and spontaneous VA has been more consistently demonstrated in larger follow-up studies and the predictive value of scar core and border zone may be comparable.6–9


Several limitations in this study need to be addressed. Inducible sustained VA during EPS is not an equivalent of SCD. Although a positive EPS is a predictor of clinical events of SCD, the predictive accuracy is limited. However, a recent study revealed that among patients with severely impaired LVEF and no inducible VA early after myocardial infarction, 93% remained free from arrhythmia or death after 3 years of follow-up without the protection of an ICD, designating its high negative predictive value.30 Second, the size of the patient population is relatively small, which may have obscured some significant imaging parameters and allowed only limited multivariable risk analyses. Hence, the results of the current study should be interpreted with caution and future studies with more relevant end points in long-term follow-up with larger number of patients are required to determine the best risk stratification approach for SCD. Third, it must be acknowledged that although global PET-assessed hyperaemic MBF appeared to be the single most important imaging parameter for the prediction of VA, the overlap between EPS-positive and EPS-negative patients was quite substantial. This implicates that clinical decision-making based on such imaging parameters will probably prove to be challenging. In addition, the frequency of ACE/ARB usage was different between EPS-positive and EPS-negative patients, which may be a confusing factor. The use of ACE/ARB, however, may indirectly result in reduced risk of VA susceptibility as these agents attenuate the ventricular remodelling process following myocardial infarction and may have sympatholytic effects. Indeed, previous studies have shown that no angiotensin inhibition is an independent predictor of SCD.12 The current study is in line with previous findings, although ACE/ARB did not remain significant in multivariable analysis. Finally, as compared to CMR, PET has some demerits for its use in clinical practice. It requires an on-site cyclotron for tracer production with short half-life such as [15O]H2O and [11C]HED, having a half-life of 2 and 20 min, respectively. Consequently, PET imaging is more costly and [15O]H2O PET or [11C]HED PET can only be performed in dedicated centres. In addition, radiation burden is of concern with PET imaging, although [15O]H2O PET imaging is accompanied by a relatively low dosimetry of ∼0.5 mSv per flow measurement.


Impairment in global hyperaemic perfusion was the only independent predictor for inducibility of VA in patients with ischaemic left ventricular dysfunction. Other previously validated approaches to evaluate the arrhythmic substrate including PET-assessed sympathetic denervation, innervation-perfusion mismatch as well as LGE-CMR-assessed scar size appeared of less significance. Moreover, a combined approach of different imaging variables did not have incremental value in predicting VA inducibility.

Key messages

What is already known on this subject?

  • Risk prediction of ventricular arrhythmias (VA) is important to improve patient selection for primary prevention implantable-cardioverter defibrillator (ICD) therapy. Non-invasive imaging of myocardial perfusion, sympathetic innervation and scar characteristics allows detailed evaluation of the arrhythmic substrate that could enhance risk stratification.

What might this study add?

  • This study compared the role of cardiovascular MRI and positron emission tomography (PET) in predicting VA inducibility in patients with ischaemic left ventricular dysfunction. Patients with inducible VA showed lower hyperaemic perfusion, larger sympathetic nerve defects and tended to have more scar. However, impaired hyperaemic perfusion was the only predictor of VA inducibility.

How might this impact on clinical practice?

  • These results suggest that impaired global hyperaemic perfusion is significantly related with electrical instability that could result in life-threatening VA, whereas sympathetic denervation and scar size appeared of less importance. Consequently, quantitative PET perfusion imaging may assist in risk stratification for VA to identify patients that will benefit most from ICD therapy.


View Abstract


  • Contributors MTR collected and analysed data, and drafted the manuscript. CPA and PK: served as scientific advisors, conceived the study protocol, collected and analysed data, and critically reviewed the manuscript. SdH: collected and analysed data, and critically reviewed the manuscript. AMB, ACvR, HJH, MCH, AAL: served as scientific advisors, conceived study protocol and critically reviewed the manuscript.

  • Competing interests None declared.

  • Patient consent Obtained.

  • Ethics approval The study was approved by the Ethics Committee of the VU University Medical Center, in accordance with the Declaration of Helsinki.

  • Provenance and peer review Not commissioned; externally peer reviewed.

Linked Articles